diff --git a/rpcs3/Emu/RSX/RSXThread.cpp b/rpcs3/Emu/RSX/RSXThread.cpp
index cd75b092f9..ed193c88e2 100644
--- a/rpcs3/Emu/RSX/RSXThread.cpp
+++ b/rpcs3/Emu/RSX/RSXThread.cpp
@@ -2334,6 +2334,12 @@ namespace rsx
 		//verify (HERE), async_tasks_pending.load() == 0;
 	}
 
+	void thread::flush_fifo()
+	{
+		// Make sure GET value is exposed before sync points
+		fifo_ctrl->sync_get();
+	}
+
 	void thread::read_barrier(u32 memory_address, u32 memory_range)
 	{
 		zcull_ctrl->read_barrier(this, memory_address, memory_range);
diff --git a/rpcs3/Emu/RSX/RSXThread.h b/rpcs3/Emu/RSX/RSXThread.h
index 75136652f7..fbc48427ea 100644
--- a/rpcs3/Emu/RSX/RSXThread.h
+++ b/rpcs3/Emu/RSX/RSXThread.h
@@ -518,6 +518,7 @@ namespace rsx
 		u32 restore_point = 0;
 		atomic_t<bool> external_interrupt_lock{ false };
 		atomic_t<bool> external_interrupt_ack{ false };
+		void flush_fifo();
 
 		// Performance approximation counters
 		struct
diff --git a/rpcs3/Emu/RSX/rsx_methods.cpp b/rpcs3/Emu/RSX/rsx_methods.cpp
index 55d55997c7..ba28a11cff 100644
--- a/rpcs3/Emu/RSX/rsx_methods.cpp
+++ b/rpcs3/Emu/RSX/rsx_methods.cpp
@@ -65,13 +65,15 @@ namespace rsx
 			rsx->sync_point_request = true;
 			const u32 addr = get_address(method_registers.semaphore_offset_406e(), method_registers.semaphore_context_dma_406e());
 
-			const auto& sema = vm::_ref<atomic_t<be_t<u32>>>(addr);
+			const auto& sema = vm::_ref<atomic_be_t<u32>>(addr);
 
 			// TODO: Remove vblank semaphore hack
-			if (sema.load() == arg || addr == rsx->ctxt_addr + 0x30) return;
+			if (sema == arg || addr == rsx->ctxt_addr + 0x30) return;
+
+			rsx->flush_fifo();
 
 			u64 start = get_system_time();
-			while (sema.load() != arg)
+			while (sema != arg)
 			{
 				if (Emu.IsStopped())
 					return;